MRO-W Reports 2008-2009

This project explores the use of machine learning techniques to aid protein engineers in the design artificial proteins. The ultimate goal of this project is synthesis of a highly active variant of the tGcn5 protein, bearing unnatural amino acids. Previous work has shown that substituting certain amino acids by unnatural fluorinated amino acids increases the stability of the protein, but inhibits the activity. We wish to find a variant of this artificial protein that remains stable but is highly active. We will use a machine-learning approach to identify variants expected to have improved activity. In particular, active learning approaches seem well suited to the problem. An initial set of tGcn5 variants has been designed. These variants will be synthesized and their activity will be measured. In parallel, several active learning algorithms will be evaluated to determine which is most suitable to the problem at hand. The selected active learning algorithm will then be used to design additional protein variants expected to have improved activity levels. These will also be evaluated experimentally.